Datacenters and cloud computing incorporate many of the same technologies and are often closely related; however, they have different functionality. Though most cloud computing centers reside in datacenter-like infrastructures, the ability to support stateless (does not keep track of configuration settings, transaction data, or other information) computing is not the role of a datacenter. In the article “Clouds are not Datacenters,” Bias describes the differences between datacenters and cloud computing (Bias 2008).
Datacenters are typically a single building specifically designed to house computer systems or telecommunications systems. Datacenters generally include redundant power and back-up power. Cloud computing is an abstraction of a facility. Typically, multiple datacenters are required to ensure cloud resources are always available. Some similarities are that both datacenters and cloud computing offer specific resources, storage, memory, and CPU power. Cloud computing users do not care where these resources come from. A datacenter is a specific location and limited in the offerings by what is located in that building. Bias explains that distributed computing is a key aspect of cloud computing which defines the relationship to application development. Development of applications destined for cloud computing requires detailed architectural forethought and detailed design work prior to fielding. Legacy standalone applications are not necessarily good candidates for implementation in a cloud environment. Application architecture designs are required before development can start. Cloud applications are specific to the environment in which they reside. Migration of an existing application to a cloud
environment typically requires added cost and time to ensure operability. The JIE strategy is an example of DOD’s plan to migrate to a cloud based architecture. This strategy is very complex. However, many commercial companies have made this same transition. The following case study of Fujitsu is an example of this transition. Fujitsu built a three-phased plan, which provided very clear guidance and execution strategy. The result was a methodical plan, which resulted in cost savings and new business markets.
Cloud Computing Case Study
Fujitsu has two different business models in which they use their cloud computing technologies: offerings to the public sector, and internal software development environments. This case study will show how a common approach helped to establish clear objectives for the company. This case study also shows that, with the right methodology, the migration to cloud computing can be profitable and can improve internal business processes. Fujitsu had three issues: increasing server operations cost, server over/under utilization, and increasing labor hours for constructing development environments.
1. Increasing server operation cost: The development centers all buy, manage, and operate their own hardware as standalone platforms.
Estimations that maintaining the servers required 1.5 person-days per month per server. As the number of servers increased, the labor hours to maintain the servers increased (Arimura and Ito 2011, 325).
2. Server over/under-utilization: Testing conducted during peak utilization periods supported the addition of more servers to meet the needs of the development teams at each location. This left under-utilization periods and inefficient usage of the resources during idle time.
3. Increasing labor hours for constructing development environments:
Middleware developers utilized independent hardware suites to support each development environment. As the number of middleware products increased, so did the number of single-focused hardware platforms to support testing. Each platform required additional resources to manage and support the hardware and software environments. With the addition of 64-bit CPUs and virtualization software, the number of platforms for testing products increased dramatically. This last problem showed an eightfold increase in the number of test platforms required between 2005 and 2009, from 10 to 84. During this same time, the labor hours needed to build a product, construct a test environment, and test the product showed
a 25% increase from 2005 to 2008 and then another 30% increase from 2008 to 2009. Because of the increase in labor hours, the cost of doing business, and the changing technologies, Fujitsu made the decision to use their own environment to support their own development (Arimura and Ito 2011, 326).
4. For three years starting in 2008, the Numazu Cloud Center undertook a conversion to a cloud-based software development environment. Fujitsu developed a three-phased methodology based on lessons learned from previous efforts. The three phases, consolidation and virtualization,
standardization, and systemization supported software development efforts as well as the commercial hosting offerings as described. Figure 5 shows the three phases and the list of objectives for each. Fujitsu used a detailed approach breaking down objectives for each phase before moving to the next phase. The case study explains how each step leads to the next but only after the objectives achieved.
Figure 5. Cloud Computing Methodology Consolidation /
Virtualization
•Consolidation
•Operational Task
•Virtualization
•Reduce Older Server Platforms / Reduce Hardware
Standardization
•Self Servicing Model / New Product Offering and Reduced Staffing
Consolidation and virtualization: during 2008, the company started the consolidation of the servers, initially targeting about 1800 servers, later realizing only about 100 really need to be consolidated. With the use of virtual servers, the middleware developers could then rent actual machines or virtual machines. Because of the nature of the development, most developers needed actual machines. Between 2008 and 2010, the number of virtual machines increased from 900 to more than 2300. This consolidation and use of virtual machines was not without challenges. The company had to add additional hardware and special software to handle the specific needs of the developers.
The result has been a more efficient development environment (Arimura and Ito 2011, 326).
Standardization: Based on analysis of a 2008 study of their virtualization environments, Arimura and Ito (2011) discovered that the company had 348 patterns, each representing different combinations of CPU number, memory size, disk capacity, and OS type. According to the same report (329), approximately 51% of the patterns found to be very similar.
Systemization: The software development partners continued to increase as other companies used the Fujitsu cloud services as their development environments. Arimura and Ito (2011) report shows the demand on resources continued to increase, placing a burden on manual processes to provision the resources needed to support the dispersed development teams. In order to meet these needs, Fujitsu developed a number of products to automate the processes. This included a service catalog of products; automated deployment environment; automated operations; dynamic resource management; and automated operations in a cloud environment.
Fujitsu offered these products to the commercial market as part of their new business line of cloud computing services. They also used the new cloud environment to re-focus their internal development teams. The lessons learned from the internal development processes helped Fujitsu to continue to improve their external commercial offerings (Arimura and Ito 2011, 329).
Fujitsu made an initial invested of $14 million in hardware and software over a three-year period. As of 2011, Fujitsu projected a continued cost reduction of $9 million annually for infrastructure and $2.5 million annually from terminating leases and consolidating services. With the deployment of the new software management tools, Fujitsu balanced utilization of servers and realize a significant reduction in labor hours to maintain and manage their cloud environments. By reducing hardware and using newer technology, Fujitsu has seen a reduction in power consumption, which added to the total life cycle savings for the company.
In summary, Fujitsu’s initial approach required modifications and development of clear goals. By creating their three-phase approach, Fujitsu established a detailed methodology to transition from standalone environments to cloud computing. Fujitsu took advantage of lessons learned from other companies that had made similar transitions. Fujitsu also expanded the lessons learned to include how they defined their internal development environment and commercial offerings.
This case study provides clear examples of how defining a clear methodical approach to cloud computing can help a company to achieve cost savings and improve business opportunities. For DOD entities, the lessons learned can help define clear methodologies that organizations can leverage when moving major applications or reducing facilities with a cloud computing initiative. Fujitsu showed that it is not only possible to reduce cost and hardware, but to expand internal development capabilities critical to business operations as well Figure 6 provides an example of a model derived from figure 5, that can be used in DOD to help program managers better manage the transition to cloud computing.
Figure 6. High Level Cloud Computing Deployment Model
Based on this simple approach, the senior levels of DOD that direct the program managers and acquisition commands to move to cloud computing could do so with a common set of standards and objectives. Technology is only one aspect of using commercial technology. DOD needs to consider the methodology that program managers will use to leverage these technologies and provide clear guidance on the objectives. The Fujitsu case study showed that by defining a three-year plan, laying out clear tasks for each year, and then empowering the program managers with execution of the plan, they were able to achieve and even exceed the initial goals. These are lessons learned that DOD should take advantage of as they move to cloud computing.
In 2010, the U.S. Chief Information Officer issued a paper titled “25 Point Implementation Plan to Reform Federal Information Technology Management” (Kundra
Methodology
2010). Many of the points in this thesis are contained in the 25-point plan. Like many management plans, focusing on how to align the money with the execution is a critical aspect of the management strategy for this effort. In the paper, the article tried to align the contract language to the Request for Proposal (RFP) process. Parts of the paper were very inspiring, and it appeared that much of the argument of the paper aligned to the execution of the management plan; however, since publication of that paper, there has been little movement toward execution of its management strategy.
Not all aspects of the Kundra (2010) plan have been lost. The consolidation of DOD datacenters by 2015 is well underway and the “Cloud First” strategy is gaining momentum across DOD. The U.S. Navy has launched initiatives to place cloud-computing devices on U.S. Navy ships. This initiative initially is only applicable on aircraft carriers because of bandwidth constraints on smaller ships today. Large deck ships may not be the classical datacenter. The amount of data that users generate or use in their daily jobs coupled with the number of applications residing on a single network, lead to the potential of a single purpose datacenter on a ship. Despite this, other aspects of the management strategy are still misaligned, such as the acquisition process and the failure to empower the program manager to field new technologies as part of their existing development. The U.S. Navy could benefit from defining a process similar to the one used by Fujitsu and shown at the high level in Figure 6.